Detecting transient brain states of functional connectivity: A comparative study

Estimating functional connectivity (FC) has become an increasingly powerful tool for understanding our brain and investigating healthy and abnormal brain functions. Most previous studies assume temporal stationarity, such as correlation and data-driven decompositions computed across the whole duration of the acquisition. However, emerging evidence revealed the presence of temporal variability of FC, leading to increasing interest in estimating the dynamic Functional Connectivity (dFC). In this context, several approaches have been proposed in order to extract the relevant brain networks fluctuating over time. Still, a clear comparative study among the existing methods is needed. Thus, we aimed here to compare three dimensionality reduction techniques, specifically Independent Component Analysis (ICA), Principle Component Analysis (PCA) and generalized Canonical Correlation Analysis (gCCA), on two Magnetoencephalography (MEG) datasets recorded during motor and memory tasks. First, source connectivity combined with a sliding window approach was used in order to reconstruct the dynamic brain networks at the cortical level. Then, for each algorithm, we extracted the significant patterns of brain network connections with their associated time variation. Results show characteristic properties of each method in terms of computation time, reproducibility and potentiality in extracting the top dominant networks.